Embryonic stromal clones reveal developmentalregulators of definitive hematopoietic stem cellsCharles Durand*†, Catherine Robin*, Karine Bollerot*, Margaret H. Baron‡, Katrin Ottersbach*, and Elaine Dzierzak*§

*Department of Cell Biology and Genetics, Erasmus Medical Center, 3000 CA, Rotterdam, The Netherlands; and ‡Departments of Medicine, Molecular, Cell,and Developmental Biology, Gene and Cell Medicine, and Oncological Sciences and The Black Family Stem Cell Institute, Mt. Sinai School of Medicine,New York, NY 10029

Edited by Paul A. Marks, Memorial Sloan–Kettering Cancer Center, New York, NY, and approved November 5, 2007 (received for review July 26, 2007)

Hematopoietic stem cell (HSC) self-renewal and differentiationis regulated by cellular and molecular interactions with the sur-rounding microenvironment. During ontogeny, the aorta–gonad–mesonephros (AGM) region autonomously generates the first HSCsand serves as the first HSC-supportive microenvironment. Becausethe molecular identity of the AGM microenvironment is as yetunclear, we examined two closely related AGM stromal clones thatdifferentially support HSCs. Expression analyses identified threeputative HSC regulatory factors, �-NGF (a neurotrophic factor),MIP-1� (a C–C chemokine family member) and Bmp4 (a TGF-�family member). We show here that these three factors, whenadded to AGM explant cultures, enhance the in vivo repopulatingability of AGM HSCs. The effects of Bmp4 on AGM HSCs werefurther studied because this factor acts at the mesodermal andprimitive erythropoietic stages in the mouse embryo. In this report,we show that enriched E11 AGM HSCs express Bmp receptors andcan be inhibited in their activity by gremlin, a Bmp antagonist.Moreover, our results reveal a focal point of Bmp4 expression inthe mesenchyme underlying HSC containing aortic clusters at E11.We suggest that Bmp4 plays a relatively late role in the regulationof HSCs as they emerge in the midgestation AGM.

aorta–gonad–mesonephros � BMP4 � microenvironment � stromal cells

The adult hematopoietic system is sustained by a cohort ofhematopoietic stem cells (HSCs) residing in specialized

niches within the bone marrow (BM) and liver at the adult andfetal stages of development. These niches provide a supportivemicroenvironment for the maintenance and growth of HSCs. Inthe marrow, HSCs are localized near the bone (1), in theendosteal niches (2, 3), and in the vascular niches of the sinusoids(4). Indeed, several signaling pathways play a role in the regu-lation of adult BM HSC activity. These include hematopoieticcytokines, intracellular regulators, and developmental regulatorssuch as the Notch, Wnt, TGF�, and Hh signaling pathways (1–3,5–7). During mouse development, the first adult-type HSCsemerge in the midgestation aorta–gonad–mesonephros (AGM)region (8). These first HSCs are generated de novo, most likelyfrom endothelial cells (possessing hemogenic potential) local-ized within the ventral wall of the dorsal aorta (9, 10). Despiteknowledge concerning the localization of these cells, little isknown about the surrounding microenvironment that initiatesand supports the growth of these first HSCs.

Attempts to study hematopoietic supportive microenviron-ments in the adult mouse initially used in vitro culture of primaryadherent BM cells (11). Subsequently, a wide variety of stromalcell lines were generated from adult BM and other hematopoi-etic territories, including the fetal liver (FL), yolk sac, and AGM(12–17). Most stromal cell lines exhibit a vascular smooth musclephenotype and, under appropriate mesenchymal lineage-differentiation culture conditions, reveal osteo-, adipo-, andchondrogenic, and/or endothelial potential (18–20). Impor-tantly, some of these stromal cell lines support HSC and hema-topoietic progenitor growth in coculture systems and, thus, arerepresentative of the in vivo microenvironment. Using previously

isolated stromal clones from the aorta and urogenital subregionsof the AGM, we found that some clones are potent supportersof HSCs (16, 17). These clones and primary cells from the AGMexhibit mesenchymal differentiation potential, and, hence, theAGM stromal cells appear to be representative of the midges-tational HSC-supportive microenvironment (19, 21).

Transcriptional profiling has provided insight into the identityof molecules elaborated by the HSC-supportive microenviron-ment. The comparative expression signatures of embryonic day(E)14 liver-derived HSC-supportive and -nonsupportive stromallines (22) show a complex genetic program involving manymolecules whose function in hematopoiesis is unknown. Thegenetic program of the developmentally earlier AGM microen-vironment was recently examined within our panel of closelyrelated AGM-derived stromal cell lines (23). In the studypresented here, further expression analysis on AGM stromal celllines identified three additional candidate HSC-regulatory mol-ecules: �-NGF, a neurotrophic factor; MIP-1�, a chemokine ofthe C–C family whose activity on HSC potential was previouslyunexplored; and Bmp4, a known developmental signaling factor.Our in vivo transplantation experiments reveal that all three ofthese factors affect HSC activity. Most interestingly, the local-ized expression of Bmp4 in the mesenchymal cells underlyingemergent clusters of hematopoietic cells in the ventral midges-tation mouse aorta strongly suggests that Bmp4 is involved in thecomplex regulation of HSC growth.

ResultsIdentification and Expression of Candidate HSC Regulators in AGMStromal Cells and Embryonic Hematopoietic Tissues. Previously,AGM stromal clones derived from E11 urogenital ridges wereshown to provide differential support for AGM HSCs in in vitrococultures: UG26.1B6 provides potent HSC support, whereasUG26.3B5 is less supportive (16, 19). To identify potential AGMHSC regulators, we compared the gene expression profile ofthese clones using arrays containing �500 genes encodingcytokines, cytokine receptors, and molecules involved in signaltransduction pathways and cell migration. The genes highlyexpressed by both stromal clones include molecules that areknown hematopoietic and developmental regulators [supportinginformation (SI) Table 3]. A few differentially expressed genes

Author contributions: C.D. and C.R. contributed equally to this work; C.D., C.R., and E.D.designed research; C.D., C.R., K.B., and K.O. performed research; M.H.B. contributed newreagents/analytic tools; C.D., C.R., K.B., K.O., and E.D. analyzed data; and C.D. and E.D.wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

†Present address: Laboratoire de Biologie du Developpement, Unite Mixte de RechercheCentre National de la Recherche Scientifique 7622, Universite Paris VI, Paris, France.

§To whom correspondence should be addressed at: Erasmus Medical Center, Departmentof Cell Biology and Genetics, P.O. Box 2040, 3000 CA, Rotterdam, The Netherlands. E-mail:e.dzierzak@erasmusmc.nl.

This article contains supporting information online at www.pnas.org/cgi/content/full/0706923105/DC1.

© 2007 by The National Academy of Sciences of the USA

20838–20843 � PNAS � December 26, 2007 � vol. 104 � no. 52 www.pnas.org�cgi�doi�10.1073�pnas.0706923105

were found, including �-NGF (a neurotrophic factor), MIP-1� (amember of the C–C chemokine family), and Bmp4 (a morpho-gen). Quantitative real-time PCR (Fig. 1) demonstrated expres-sion of �-NGF and Bmp4 at 2.5- and 4-fold higher levelsrespectively in UG26.1B6 cells, whereas MIP-1� was expressedat a 5-fold higher level in UG26.3B5 cells. Thus, �-NGF andBmp4 may act as positive regulators, and MIP-1� may act as anegative regulator of AGM HSC growth.

To determine whether the candidate regulators are normallyexpressed within embryonic hematopoietic sites, RT-PCR for�-NGF, MIP-1�, and Bmp4 and their receptors was performedon aorta mesenchyme, urogenital ridges, liver, yolk sac, and limbbud (positive control) tissues from E11 mouse embryos (Fig. 2).�-NGF is expressed at low levels in both subregions of the AGMbut not the liver or yolk sac. Whereas the p75 receptor gene isexpressed in aorta, urogenital ridges, and liver, the high-affinityreceptor gene TrkA is expressed only in urogenital ridges andliver. MIP-1� and its receptor Ccr1 are expressed in bothsubregions of the AGM and the other hematopoietic tissues.Bmp4 and Bmp receptors Alk3, Alk6, and BmprII are expressedin all of the embryonic hematopoietic tissues, as are the intra-

cellular signaling molecules Smad1, 4, and 5 (the latter only atlow levels). These results suggest that the three regulatorypathways may be active in vivo in the AGM region and couldaffect HSC development.

Effects of Recombinant �-NGF, MIP-1� and Bmp4 on AGM HSCs.Previously, others have suggested that �-NGF and its receptorTrkA participate in the control of hematopoiesis (24, 25). AMIP-1� family member MIP-1�, which shares 24% sequencehomology, is a known inhibitor of BM HSC proliferation (26,27). Bmp4 is known to affect the commitment of mesoderm tohematopoietic fate in embryos (28, 29) and ES cells (30, 31),primitive erythropoiesis (32), and the expansion of humanhematopoietic progenitors (33). Thus, to study the effects of�-NGF, MIP-1�, and Bmp4 on AGM hematopoiesis and HSCs,we performed in vivo transplantation experiments of cells fromE11 AGMs after 3 days of explant culture in the presence ofthese exogenous factors. This organ culture system has beenshown to maintain and promote the autonomous expansion ofAGM HSCs and CFU-S (8).

As shown in Table 1, HSC activity is increased almost 2-foldin the presence of 250 ng/ml �-NGF (60% of recipients repop-ulated) compared with AGM explants cultured in its absence(36% of recipients repopulated) or with 25 ng/ml �-NGF. A1.3-fold increase in HSC activity was found when AGM explantswere cultured with 20 ng/ml Bmp4 (77% of recipients repopu-lated compared with control 58%). No increase in HSC activitywas found in the presence of 2 ng/ml Bmp4 (43% of recipientsrepopulated). These factor-induced increases in HSC activitysuggest that the higher expression of �-NGF and Bmp4 byUG26.1B6 (compared with UG26.3B5) may be responsible, inpart, for its in vitro support of HSCs.

In the presence of both 20 ng/ml and 200 ng/ml MIP-1�, AGMHSC activity was also increased (1.5- to 2-fold). This is incontrast to the finding that the UG26.3B5 cells, which are lesssupportive for BM HSCs, express much higher levels of MIP-1�than UG26.1B6 cells. Because others have found that MIP-1�inhibits adult BM HSC proliferation, we tested whether MIP-1�would increase or decrease adult BM HSC activity. WhenMIP-1� was added in cocultures of sorted c-kit�Ly-6A GFP�

BM cells and UG26.1B6, HSC activity was diminished (Table 2).Thus, MIP-1� acts similarly to family member MIP-1� as anegative regulator of adult BM HSCs. The observation thatMIP-1� acts as a positive regulator of AGM HSCs suggests thatthis factor acts differentially in the embryo and adult.

Bmp4 Regulates AGM HSCs. Considering that �-NGF is not highlyexpressed in the aorta subregion and that MIP1-� effects onAGM HSCs are complex, we focused further studies on Bmp4.To date, no studies have addressed the functional role of Bmp4in the regulation of the first HSCs in the AGM microenviron-ment. If Bmp4 does play a role, it should be possible to modulateHSC activity in AGM explants by inhibiting the Bmp4 pathwaywith gremlin, an antagonist known to bind Bmps and block Bmpsignaling (34).

To examine the role of Bmp4 produced endogenously withinthe AGM, we added recombinant gremlin to explant cultures.Gremlin activity (250 ng/ml) was verified by its inhibition ofBmp4-mediated differentiation of an osteogenic cell line (SI Fig.5). When gremlin was added to AGM explants (250 ng/ml), HSCactivity was abolished (Fig. 3A). Noggin (250 ng/ml), anotherBmp antagonist, also reduced AGM HSC activity, although to alesser extent (data not shown). The effect of Bmp4 on immaturehematopoietic progenitors was also tested by the short-term invivo Colony Forming Unit–Spleen (CFU-S) and in vitro ColonyForming Unit-Culture (CFU-C) assays. E11 AGM explantscultured in the presence of gremlin yielded a statistically signif-icant decrease (3-fold) in the number of CFU-S11 (Fig. 3B) and

Fig. 1. Differential expression of candidate HSC regulators by UG26.1B6 andUG26.3B5 stromal cells. Real-time PCR was performed for �-NGF, MIP-1�, andBmp4 on three RNA samples from UG26.1B6 (HSC-supportive) and UG26.3B5(less supportive) stromal clones. Graphs show the relative expression levels ofeach factor in both cell lines. n � 3. For MIP-1� and Bmp4, P � 0.001.

Fig. 2. Expression of candidate regulators and receptors in E11 embryonictissues and enriched AGM HSCs. RT-PCR analysis of E11 embryonic tissues forexpression of �-NGF and NGF receptors (p75 and TrkA), MIP-1� and receptorCcr1, Bmp4, and Bmp4 receptors (Alk3, Alk6, and Bmp-RII), and downstreamsignaling molecules Smad1, 4, and 5 (A); GFP� and GFP� sorted AGM cells fromE11 Ly-6A GFP embryos for expression of Alk3, Alk6 and BmprII (B); and forexpression of Bmp4 and Ccr1 (C). �-actin expression was the normalizationcontrol. Ao, aorta and mesenchyme; UG, urogenital ridges; FL, fetal liver; YS,yolk sac; LB, limb bud.�rt, no reverse transcriptase.

Durand et al. PNAS � December 26, 2007 � vol. 104 � no. 52 � 20839

DEV

ELO

PMEN

TAL

BIO

LOG

Y

CFU-Mix (Fig. 3C; 2-fold decrease) per AGM (P � 0.01). Incontrast, gremlin did not significantly affect the number of moremature progenitors (CFU-G, CFU-M, and CFU-GM). Takentogether, Bmp4 signaling appears to be involved in positivelyregulating the number of the most immature AGM hematopoi-etic progenitors and HSCs.

To investigate how Bmp4 mediates its effect on AGM hema-topoiesis, the viability of AGM hematopoietic cells cultured inthe presence of gremlin was tested. Flow cytometric analysisshowed no difference in the percentage and absolute number ofCD45� cells in E11 AGM explants cultured with or withoutgremlin (Fig. 3 D and E). When these cells were analyzed for7AAD uptake and expression of the apoptotic marker AnnexinV, no differences in the percentages and absolute numbers ofCD45� cells in the preapoptotic or apoptotic stages were ob-served (Fig. 3 D and E). Therefore, high-dose gremlin is not toxicand does not compromise AGM hematopoietic cell viability orpredispose these cells to the apoptotic pathway.

We next examined the effects of gremlin and Bmp4 onendothelial (CD45�Flk-1�CD34�) and hematopoietic progeni-tor/stem cell (CD34�c-kit�) populations in AGM explants (Fig.3F). Gremlin affected a small increase and Bmp4 a decrease inthe percentages of CD45�Flk-1�CD34� cells in AGM explants.However, no obvious differences were found for the CD34�c-kit� population cultured with either gremlin or Bmp4. Thesedata indicate that Bmp4 signaling is not involved in the survivalof preexisting HSCs or hematopoietic progenitors and, togetherwith the in vivo functional results, suggest that Bmp signalingplays a role in maintaining HSC and hematopoietic progenitorpotential.

Localized Expression of Bmp4 in the E11 AGM Region. AlthoughMarshall et al. (35) showed a polarized gradient of Bmp4 proteinunderlying the ventral wall of the human aorta at week 5 ofgestation, no comparable study has been performed in themouse E11 AGM, at the time when HSC activity is highest.Bmp4 immunostaining analysis of an E11 mouse trunk section

Table 1. Effects of recombinant factors on AGM HSC repopulating activity

Recombinant factor Cells injectedRepopulated/transplanted

(mice repopulated, %)Donor chimerismrange (mean, %)

None 0.3 ee 8/22 (36) 27–100 (72)�-NGF (25 ng/ml) 0.3 ee 3/8 (38) 29–76 (58)�-NGF (250 ng/ml) 0.3 ee 6/10 (60) 25–100 (75)MIP-1� (20 ng/ml) 0.3 ee 8/13 (62) 16–100 (75)MIP-1� (200 ng/ml) 0.3 ee 8/15 (53) 22–100 (74)None 1 ee 7/12 (58) 37–77 (57)Bmp4 (2 ng/ml) 1 ee 6/14 (43) 23–72 (46)Bmp4 (20 ng/ml) 1 ee 10/13 (77) 21–69 (48)

E11 AGM explants were cultured in the presence of recombinant factors. After 3 days of culture, AGM cells were collected, and 0.3 or 1 embryo equivalents(ee) of cells were injected into irradiated adult recipient mice. Four months after transplantation, mice were analyzed (�10% of donor chimerism is consideredpositive engraftment). n � 3 for MIP-1� and Bmp4 and n � 1 for �-NGF.

Table 2. Effects of recombinant MIP-1� on enriched BM HSCrepopulating activity

Recombinantprotein

Cell doseinjected

Repopulated/transplanted(mice repopulated, %)

None 5.0 � 103 3/6 (50)MIP-1� (20 ng/ml) 5.0 � 103 1/6 (15)MIP-1� (200 ng/ml) 5.0 � 103 0/6 (0)None 2.5 � 103 2/6 (30)MIP-1� (20 ng/ml) 2.5 � 103 2/6 (30)MIP-1� (200 ng/ml) 2.5 � 103 0/6 (0)

Adult BM c-kit�Ly-6A GFP� cells were sorted and cocultured with UG26.1B6stromal cells in the presence of MIP-1�. After 10–11 days, all cells wereharvested, and the equivalent of 2.5 or 5.0 � 103 input c-kit�Ly-6A GFP� cellswere injected into irradiated adult mouse recipients. Four months aftertransplantation, mice were analyzed (�10% of donor chimerism is consideredpositive engraftment). Chimerism ranged from 10% to 100%. n � 2.

Fig. 3. Effect of gremlin on the activity and survival of AGM hematopoieticcells. (A–C) E11 AGMs cultured as explants in the presence or absence ofgremlin (250 ng/ml) were analyzed after long term HSC repopulation assay(A), CFU-S11 assay (B), and CFU-C assay (C). For HSC repopulation, 0.3 AGM cellequivalents were injected per recipient. For each CFU-S11 condition, 14 micewere injected with 1 AGM equivalent (n � 3, *, P � 0.002). For CFU-C, n � 3.P � 0.01 for CFU-Mix. Data represent the mean � SD. (D) FACS plots showing7AAD and Annexin V staining of CD45� cells from AGMs cultured in theabsence or presence of gremlin (250 ng/ml). Percentages of cells shown ingated regions. (E) Graphs show absolute numbers of CD45� cells and AnnexinV� 7AAD�CD45� cells per AGM (n � 3). (F) Four-color staining dot plotsshowing percentages of viable cells from AGM explants cultured in thepresence or absence of gremlin or Bmp4. (Upper) Cells gated within the CD45�

population. (Lower) Cells gated within the total population. C, control; G,gremlin.

20840 � www.pnas.org�cgi�doi�10.1073�pnas.0706923105 Durand et al.

(Fig. 4A) shows strong localized Bmp4 expression ventral to thedorsal aorta. A patchy signal is seen in areas surrounding thenotochord, mesonephros, the region between the dorsal aortaand gut, and dorsal to the cardinal veins. Ventral to the aorta,Bmp4 is found mainly in mesenchymal cells and appearsthroughout the cell cytoplasm (Fig. 4B). A control stainedsection (Fig. 4C) and adjacent anti-Bmp4 stained section (Fig.

4D) verified the specificity of staining. To determine whetherthese cells are actively transcribing Bmp4, in situ analysis wasperformed. Bmp4 expression was found in cells in the samemesenchymal region underlying the ventral aspect of the dorsalaorta (Fig. 4 E and G).

Immunostaining for Bmp4 was also performed on sectionsfrom E11 transgenic Ly-6A GFP embryos to test whetherBmp4-expressing cells are associated with emerging HSCs in thecell clusters along the ventral luminal wall of the dorsal aorta. Ithas been shown that AGM HSCs are exclusively found in theGFP� fraction (9). A concentrated localization of Bmp4 expres-sion was found in mesenchymal cells underlying GFP� cells inthe ventral endothelial wall (Fig. 4F). Moreover, whereas themesenchymal cells are positive for Bmp4 throughout the cyto-plasm, some cells, particularly the GFP� cells in the hemato-poietic clusters (Fig. 4H), appear to have foci of Bmp4 on theirsurface suggestive of an interaction with surface receptors.

Because Bmp4 receptors are known to be expressed by adultBM HSCs (33), we tested whether they are also expressed onAGM HSCs. A enriched HSC population (GFP�) was sortedfrom Ly-6A GFP transgenic E11 aortas and analyzed by RT-PCRfor expression of Bmp receptors Alk3, Alk6, and BmprII. GFP�

cells express Alk3 and BmprII but not Alk6 (Fig. 2B), whereas allthree receptors were expressed by GFP� cells. RT-PCR analysiswas also performed for Bmp4. As expected, Bmp4 is highlyexpressed in GFP� cells. GFP� cells express only low levels ofBmp4 (Fig. 2C). Taken together, these data suggest that Bmp4produced by the mesenchymal cells underlying the aortic hema-topoietic clusters directly influences AGM HSC development.

DiscussionIdentification of Regulatory Molecules Expressed by the AGM Hema-topoietic Supportive Microenvironment. Gene expression profilingof HSC-supportive and -nonsupportive cell lines as first reportedby Moore and colleagues (22) has provided insight into themolecular players in the fetal/adult hematopoietic microenvi-ronment. Our studies focused on the AGM microenvironment(23) with the UG26.1B6 clone, the most highly HSC-supportiveAGM stromal line (16). It was derived from E11 urogenital(UG) tissue along with a closely related stromal line that doesnot efficiently support HSCs (UG26.3B5). Although it wasinitially surprising to find HSC-supportive cells in the UGsubregion of the AGM, HSCs have been found in E11 UGs afterexplant culture and in vivo in E12 UGs (9). Similar to bonemarrow-derived stromal cells, UG26.1B6 express the Sca-1marker (17). Sca-1 is expressed at high levels in the UG regionof E11 embryos (36), strongly suggesting that UG26.1B6 repre-sents a component of the in vivo AGM HSC niche.

�-NGF, MIP-1�, and Bmp4, identified through their differ-ential expression in the UG stromal clones, are able to affectAGM HSC growth in vitro. Both �-NGF and Bmp4 were morehighly expressed in supportive UG26.1B6 cells than in nonsup-portive UG26.3B5 cells. As expected for positive regulators,addition of exogenous �-NGF or Bmp4 to AGM explantsresulted in approximately a 2-fold increase in HSCs (as deter-mined by in vivo transplantation). In contrast, MIP-1� wasexpressed at higher levels by the nonsupportive cell lineUG26.3B5. When MIP-1� was exogenously added to AGMexplants, it did not inhibit HSCs but, instead, showed a positiveeffect. Hence, expression studies must be complemented byfunctional assays.

The stimulatory effect of MIP-1� on AGM HSCs was sur-prising, given that this factor inhibited adult BM HSCs. We havefound by RT-PCR that BM c-kit�Ly-6A GFP� cells express theMIP-1� receptor Ccr1 (data not shown), suggesting that BMHSCs can directly respond to MIP-1�. The effect of MIP-1� onAGM HSCs appears to be more complex, because both Ly-6AGFP� and GFP� AGM cells express Ccr1 (Fig. 2C). It is possible

Fig. 4. Expression of Bmp4 in the midgestation AGM. Transverse sectionsthrough E11 AGM (dorsal at top; ventral at bottom) stained (A) with anti-Bmp4 antibody. (B) Framed region in A showing Bmp4 staining (green) in themesenchyme underlying the ventral wall of the dorsal aorta (arrowheads).Nuclear DAPI staining is blue. (C) Immunostaining control with fluorescent(red) secondary antibody used in D in combination with anti-Bmp4 antibody(late E10/early E11). (E) In situ hybridization analysis showing Bmp4 mRNA(purple). Expression is restricted to mesenchyme underlying the ventral aspectof the dorsal aorta (arrowheads). (F) Transverse section through E11 Ly-6A GFPaorta stained with anti-Bmp4 antibody. GFP� cells (green) are located in theventral endothelium of the aorta with Bmp4 protein (red) detected in themesenchyme underlying GFP� cells (arrowheads). (G) Framed region in Eshowing Bmp4-expressing cells (arrowheads) underlying a hematopoieticcluster (arrow). (H) Ventral aspect of aorta in F showing Bmp4 in mesenchymalcells (arrowhead) underlying GFP� hematopoietic cells (arrow). Some GFP�

cells show punctate Bmp4 signal. cv, cardinal vein; da, dorsal aorta; e, eryth-rocytes; g, gonad; m, mesonephros; n, notochord.

Durand et al. PNAS � December 26, 2007 � vol. 104 � no. 52 � 20841

DEV

ELO

PMEN

TAL

BIO

LOG

Y

that MIP-1� directly affects HSCs. Alternatively or additionallyit could act through nonhematopoietic cells in the microenvi-ronment and they, in turn affect HSCs. UG26.1B6 cells wouldthen represent only a part of the AGM microenvironment.Because we have found that Ccr1 is expressed by both the aorticand UG regions of the AGM, identification of the respondingcells will be of great interest. Notwithstanding, MIP-1� is one ofthe few factors identified to date that differentially regulatesadult and embryonic HSCs.

It is intriguing that the well studied neuronal growth factor�-NGF is expressed by AGM stromal lines and increases AGMHSC repopulating activity. Parallels between neuronal and hema-topoietic development have been suggested by several studies (37),particularly in the formation of the enteric nervous system and thelymphoid Peyer’s patches (38). Both require signaling through thesame receptor system, the RET receptor tyrosine kinase andGfr�-3, a member of the GDNF family of neuronal receptors.Moreover, many neuronal factors were found to be expressed byfetal liver stromal lines, including neurotactin, neuromodulin, neu-roligin2, neuregulin, and pro-NDF-�1 (22). Thus, factors tradition-ally associated with one lineage may also regulate the developmentof a different lineage. Further studies with neuronal factors will beparticularly interesting because, simultaneous to the generation ofHSCs, neural crest cells are migrating through the AGM region ofthe mouse embryo (39).

Bmp4 and Its Role in the Growth of AGM HSCs. Several laboratorieshave shown that the Bmp signaling pathway, and more specifi-cally Bmp4, is closely involved in the commitment of mesodermalcells to the hematopoietic lineage in both Xenopus and mammals(29), plays a role in the hematopoietic differentiation of mouseES cells (40) and controls the expansion of human BM immatureprogenitors (33). Our findings of high-level Bmp4 expression byan HSC-supportive AGM stromal cell line, the localized expressionof Bmp4 in mouse AGMs, and the results of transplantation studiesimplicate Bmp4 as a positive regulator of AGM HSCs in vivo.

The response of AGM HSCs to Bmp4 appears to be dose-dependent. Addition of exogenous Bmp4 to AGM cultures at 20ng/ml resulted in an increase in HSC activity. Addition of 200ng/ml Bmp4 did not further increase the HSC activity (data notshown), whereas 2 ng/ml had no (or a slightly negative effect) onAGM HSCs compared with the control. Bhatia and colleagues(33) found that human cord blood CD34�CD38�Lin� cellscultured with a similar medium dose of BMP4 (25 ng/ml)increased CFU-C numbers and the frequency of engraftment inNOD/SCID recipients. A lower dose of Bmp4 (5 ng/ml) resultedin a slight decrease in CFC numbers and a much lower frequencyof engraftment. Interestingly, the addition of Bmp antagonistgremlin had a more dramatic effect on AGM HSCs. Gremlin(250 ng/ml) abolished AGM HSC activity and decreased CFU-Sand CFU-Mix numbers. This was not due to a toxic effect,because high-dose gremlin did not decrease the absolute num-bers of hematopoietic cells in AGM explants nor did it induceapoptosis. Preliminary flow-cytometric analytical data have re-vealed that the numbers of CD45�CD34intermediatec-kit� cells areincreased and the numbers of CD45�CD34�Flk1� cells aredecreased in AGM explants cultured in the presence of Bmp4,suggesting that Bmp4 may induce hematopoietic fate, particu-larly to immature hematopoietic progenitors and HSCs in theaortic clusters. Because Bmp4 is locally produced in AGMmesenchymal cells adjacent to aortic endothelium and hemato-poietic clusters and Bmp4 receptors Alk3 (Type I) and BmpRII(Type II) are expressed by AGM HSCs, we suggest that endog-enous Bmp4 regulates HSC development in the AGM in vivo.Future studies will focus on demonstrating the in situ activationof the Bmp4 signaling pathway and the downstream events inspecific subpopulations of AGM cells and possible in vivo roles

of antagonists, e.g., gremlin, noggin, and follistatin and otherBmps, such as Bmp2 and Bmp7 (29) in regulating HSCs.

In conclusion, the molecular analysis of two closely relatedAGM stromal clones has allowed the identification and func-tional characterization of three regulators (MIP-1�, �-NGF, andBmp4) and one known regulator (Bmp4) of hematopoietic cells.Our studies suggest that the signaling pathways stimulated bythese factors are functional in the midgestation AGM andprovide positive signals to increase AGM HSC activity. More-over, Bmp4 does not act as a survival factor but may promote thedifferentiation and/or expansion of AGM HSCs. Future studiesidentifying the specific target populations of Bmp4 signaling (aswell as those of MIP-1� and �-NGF), such as AGM HSCs,hemangioblasts, or hemogenic endothelial cells, could provideinsights into how HSCs are generated and/or expanded inembryonic and perhaps adult vascular and mesenchymal micro-environments.

Materials and MethodsAnimals. Embryos were generated from human �-globin (41), Ly-6A GFPtransgenic (9), and C57BL/6 or (B10xCBA)F1 mice. The day of vaginal pluggingis day 0. BM was from adult Ly-6A GFP mice. Mice were bred at ErasmusMedical Center according to institutional guidelines and procedures carriedout in compliance with the Standards for Humane Care and Use of LaboratoryAnimals.

Expression Analyses. RNA was prepared from TRIzol-lysed cells (Life Technol-ogies), RNase-free DNaseI (Promega) treated, extracted in phenol/chloroform/isoamyl alcohol, ethanol precipitated, and quantified by spectrophotometry.RNA integrity was confirmed by gel electrophoresis.

Total RNA (2 �g) from UG26.1B6 and UG26.3B5 cells was used for mousecytokine array screening (R & D Systems). Gene expression was quantitated byPhosphorImaging and autoradiography (X-O Mat, Kodak) by using Image-Quant 5.2 (Molecular Dynamics). Confirmation of differential expression wasby real-time PCR. cDNA samples were amplified on an Icycler thermal cycler(Bio-Rad) by using SYBR green fluorescence and analyzed with Icycler IQdetection system software. GAPDH was used to normalize cDNA dilutions.After amplification of serial dilutions of cDNAs, cycle threshold was plottedagainst cDNA amount, and the slope was used to calculate PCR efficiency.Quantitation of samples was by comparison of the number of cycles requiredto reach reference and target threshold values.

Tissue-specific gene expression was analyzed by RT-PCR on total RNA frommouse embryonic tissues and sorted AGM cells. OligodT primers (Promega)and reverse transcriptase (SuperscriptII; Stratagene) were used for the cDNAsynthesis. Specific primers are listed in SI Table 4.

Hematopoietic Assays. Recombinant factors Bmp4, gremlin, MIP-1�, and NGFwere from R & D Systems.

In Vitro Cultures. E11 AGMs were dissected and cultured as explants on0.65-�m filters at the liquid–air interface in long-term medium (M5300; StemCell Tech) with 1 �M hydrocortisone (42). After 3 days at 37°C, explants weredissociated in collagenase. Adult BM cells were sorted for c-kit and Ly-6A GFPexpression. Sorted cells (15� 103 to 25 � 103 cells per well) were coculturedwith confluent irradiated (30 Gy) stromal cells in M5300 medium with 1 �Mhydrocortisone at 33°C, 5% CO2. Medium (half) was changed after 1 week ofculture. After 10–11 days, adherent and nonadherent cells were assayed.

In Vivo Transplantation Assay. Cells harvested from collagenase-treated ex-plants or pooled cells from BM cocultures (2,500 and 5,000 input cells) wereinjected intravenously into irradiated recipients (9 Gy). Syngeneic spleen cells(2 � 105) were coinjected to confer short-term survival of recipient. Fourmonths after transplantation, peripheral blood DNA was analyzed by semi-quantitative PCR for the donor marker (male Ymt, h�-globin, or GFP). Recip-ients were considered positive only if �10% of peripheral blood DNA was ofdonor genotype.

CFU-S11 Assay. AGM cells were injected intravenously into irradiated recipients(10 Gy) (43) and analyzed 11 days after transplantation. Spleens were fixed inTellesnicky’s solution and CFU-S counted under a dissection microscope.

20842 � www.pnas.org�cgi�doi�10.1073�pnas.0706923105 Durand et al.

CFU-C Assay. Cells were seeded in methylcellulose medium (StemCell Tech)supplemented with SCF, IL-3, IL-6, and Epo, incubated at 37°C in 5% CO2.CFU-G (granulocyte), M (macrophage), GM (granulocyte-macrophage), Mix(granulocyte, erythroid, macrophage, and megakaryocyte) were scored withan inverted microscope at day 10.

Immunostaining and in Situ Hybridization. E11 Ly-6A GFP and wild-type em-bryos were fixed in 4% PFA, embedded in PBS/gelatin/sucrose, quick-frozen,and cut at 10 or 20 �m. For Bmp4 protein detection, frozen sections werepermeabilized in PBS/0.1% Triton X-100. Immunostainings were as described(44) by using goat anti-human Bmp4 primary antibody (1:50; Santa CruzBiotechnology) and HRP- (DAKO) or biotin-(Vector Laboratories) conjugatedrabbit anti-goat secondary antibodies with streptavidin-FITC (PharMingen) or-Cy3 (Jackson ImmunoResearch).

In situ hybridization used an anti-sense riboprobe to full-length murineBmp4 cDNA (kind gift of B. Hogan, Duke University, Durham, NC) labeled byusing in vitro transcription kit (Promega) and Digoxigenin-11-UTP (Roche).Frozen sections were incubated overnight at 65.5°C in hybridization mix (1�salt, 50% formamide, 10% dextran sulfate, 1 mg/ml yeast RNA, 1� Denhardt’s)with Bmp4 probe. Slides were washed at 65.5°C in 1� SSC/50%formamide/0.1%Tween-20 and then in PBS at room temperature (RT). Sections wereblocked in PBS/10% heat-inactivated FCS/0.1% Triton X-100 at RT and incu-bated overnight at 4°C with alkaline phosphatase-conjugated anti-digoxigenin antibody (1:2000, Roche). Slides were washed in PBS/0.1% Triton

X-100 and incubated in 2% NBT/BCIP (Roche) in 0.1 M Tris, pH 9.5/0.1 M NaClseveral hours at 37°C.

Flow-Cytometric Analysis. Antibodies (BD Pharmingen) included: FITC-anti-c-kit, PE-anti-Flk1, APC-anti-CD34, FITC-anti-AnnexinV, phycoerythrin (PE), andPerCP-Cy5-anti-CD45. After 20 min of antibody incubation, cells were washedin PBS/10%FCS/PS and stained with 7AAD or Hoechst 33258 (1 �g/ml, Molec-ular Probes). Analysis was on a FACScan or FACS ARIA (Becton Dickinson). ForAnnexinV staining, cells were incubated 15 min with FITC-anti-AnnexinV and7AAD and with PE-anti-CD45.

Statistical Analysis. Student’s t test was used to determine statistical signifi-cance. P values �0.05 were considered significant.

ACKNOWLEDGMENTS. We thank all laboratory members for constructivediscussions; K. van der Horn, F. Wallberg, and R. van der Linden for cell sorting;F. Cerisoli for RNA isolations; and Dr. S. Philipsen for critical reading of themanuscript. M.H.B. thanks D. Mohn for technical assistance. This work wassupported by l’ARC (C.D.), National Institutes of Health Grants R37DK51077(to E.D.) and R01 DK52191 (to M.H.B.), Netherlands Besluit Subsidies Invest-eringen Kennisinfrastructuur Stem Cells in Development and Disease Grant03038 (to E.D.), Nederlandse Organisatie voor Welenschappelijk OnderzockVici Grant 916.36.601 (to E.D.), Marie Curie Grants HPMF-CT-2000-00871 (toC.R.), and MEIF-CT-2006-025333 (to K.B.).

1. Taichman RS, Reilly MJ, Emerson SG (2000) Hematology 4:421–426.2. Zhang J, Niu C, Ye L, Huang H, He X, Tong WG, Ross J, Haug J, Johnson T, Feng JQ, et

al. (2003) Nature 425:836–841.3. Calvi LM, Adams GB, Weibrecht KW, Weber JM, Olson DP, Knight MC, Martin RP,

Schipani E, Divieti P, Bringhurst FR, et al. (2003) Nature 425:841–846.4. Kiel MJ, Yilmaz OH, Iwashita T, Terhorst C, Morrison SJ (2005) Cell 121:1109–1121.5. Duncan AW, Rattis FM, DiMascio LN, Congdon KL, Pazianos G, Zhao C, Yoon K, Cook

JM, Willert K, Gaiano N, Reya T (2005) Nat Immunol 6:314–322.6. Reya T, Duncan AW, Ailles L, Domen J, Scherer DC, Willert K, Hintz L, Nusse R, Weissman

IL (2003) Nature 423:409–414.7. Bhardwaj G, Murdoch B, Wu D, Baker DP, Williams KP, Chadwick K, Ling LE, Karanu FN,

Bhatia M (2001) Nat Immunol 2:172–180.8. Medvinsky A, Dzierzak E (1996) Cell 86:897–906.9. de Bruijn MF, Ma X, Robin C, Ottersbach K, Sanchez MJ, Dzierzak E (2002) Immunity

16:673–683.10. North T, de Bruijn M, Stacy T, Talebian L, Lind E, Robin C, Binder M, Dzierzak E, Speck

N (2002) Immunity 16:661–672.11. Dexter TM (1979) Acta Haematol 62:299–305.12. Moore KA, Ema H, Lemischka IR (1997) Blood 89:4337–4347.13. Yoder MC, Papaioannou VE, Breitfeld PP, Williams DA (1994) Blood 83:2436–2443.14. Ohneda O, Fennie C, Zheng Z, Donahue C, La H, Villacorta R, Cairns B, Lasky LA (1998)

Blood 92:908–919.15. Xu MJ, Tsuji K, Ueda T, Mukouyama YS, Hara T, Yang FC, Ebihara Y, Matsuoka S,

Manabe A, Kikuchi A, et al. (1998) Blood 92:2032–2040.16. Oostendorp RA, Harvey KN, Kusadasi N, de Bruijn MF, Saris C, Ploemacher RE, Med-

vinsky AL, Dzierzak EA (2002) Blood 99:1183–1189.17. Oostendorp RA, Medvinsky AJ, Kusadasi N, Nakayama N, Harvey K, Orelio C, Otters-

bach K, Covey T, Ploemacher RE, Saris C, Dzierzak E (2002) J Cell Sci 115:2099–2108.18. Charbord P, Oostendorp R, Pang W, Herault O, Noel F, Tsuji T, Dzierzak E, Peault B

(2002) Exp Hematol 30:1202.19. Durand C, Robin C, Dzierzak E (2006) Haematologica 91:1172–1179.20. Chateauvieux S, Ichante JL, Delorme B, Frouin V, Pietu G, Langonne A, Gallay N,

Sensebe L, Martin MT, Moore KA, Charbord P (2007) Physiol Genomics 29:128–138.21. Mendes SC, Robin C, Dzierzak E (2005) Development 132:1127–1136.22. Hackney JA, Charbord P, Brunk BP, Stoeckert CJ, Lemischka IR, Moore KA (2002) Proc

Natl Acad Sci USA 99:13061–13066.

23. Oostendorp RA, Robin C, Steinhoff C, Marz S, Brauer R, Nuber UA, Dzierzak EA, PeschelC (2005) Stem Cells 23:842–851.

24. Chevalier S, Praloran V, Smith C, MacGrogan D, Ip NY, Yancopoulos GD, Brachet P,Pouplard A, Gascan H (1994) Blood 83:1479–1485.

25. Auffray I, Chevalier S, Froger J, Izac B, Vainchenker W, Gascan H, Coulombel L (1996)Blood 88:1608–1618.

26. Graham GJ, Wright EG, Hewick R, Wolpe SD, Wilkie NM, Donaldson D, Lorimore S,Pragnell IB (1990) Nature 344:442–444.

27. Dunlop DJ, Wright EG, Lorimore S, Graham GJ, Holyoake T, Kerr DJ, Wolpe SD, PragnellIB (1992) Blood 79:2221–2225.

28. Winnier G, Blessing M, Labosky PA, Hogan BL (1995) Genes Dev 9:2105–2116.29. Snyder A, Fraser ST, Baron MH (2004) J Cell Biochem 93:224–232.30. Choi K, Kennedy M, Kazarov A, Papadimitriou JC, Keller G (1998) Development

125:725–732.31. Zafonte BT, Liu S, Lynch-Kattman M, Torregroza I, Benvenuto L, Kennedy M, Keller G,

Evans T (2007) Blood 109:516–523.32. Schmerer M, Evans T (2003) Blood 102:3196–3205.33. Bhatia M, Bonnet D, Wu D, Murdoch B, Wrana J, Gallacher L, Dick JE (1999) J Exp Med

189:1139–1148.34. Balemans W, Van Hul W (2002) Dev Biol 250:231–250.35. Marshall CJ, Kinnon C, Thrasher AJ (2000) Blood 96:1591–1593.36. Miles C, Sanchez MJ, Sinclair A, Dzierzak E (1997) Development 124:537–547.37. Terskikh AV, Easterday MC, Li L, Hood L, Kornblum HI, Geschwind DH, Weissman IL

(2001) Proc Natl Acad Sci USA 98:7934–7939.38. Veiga-Fernandes H, Coles MC, Foster KE, Patel A, Williams A, Natarajan D, Barlow A,

Pachnis V, Kioussis D (2007) Nature 446:547–551.39. Gammill LS, Gonzalez C, Gu C, Bronner-Fraser M (2006) Development 133:99–106.40. Adelman CA, Chattopadhyay S, Bieker JJ (2002) Development 129:539–549.41. Strouboulis J, Dillon N, Grosveld F (1992) Genes Dev 6:1857–1864.42. Robin C, Dzierzak E (2005) in Developmental Hematopoiesis Methods and Protocols,

ed Baron MH (Humana, Totowa, NJ), Vol 105, pp 257–272.43. Medvinsky AL, Samoylina NL, Muller AM, Dzierzak EA (1993) Nature 364:64–67.44. Jaffredo T, Gautier R, Eichmann A, Dieterlen-Lievre F (1998) Development 125:4575–

4583.

Durand et al. PNAS � December 26, 2007 � vol. 104 � no. 52 � 20843

DEV

ELO

PMEN

TAL

BIO

LOG

Y